Factored models for Deep Machine Translation
Kiril Simov, Iliana Simova, Velislava Todorova, Petya Osenova · Bulgarian Portal for Open Science · 2015
In this paper, we present some preliminary results on Statistical Machine Translation from Bulgarian-to-English and English-to-Bulgarian. Linguistic knowledge has been added gradually as factors in the MOSES system. The tests were performed on the QTLeap corpus data in IT domain for Pilot 1. The training was done on news parallel data as well as on IT domain data. The BLEU scores show that the addition of linguistic knowledge improves the Machine Translation.